The malaria parasite Plasmodium falciparum kills hundreds of thousands of people every year. But for the majority of this parasite’s proteins we have no idea what they do and how they contribute to the parasite's ability to infect and persist in the human host. Genome-wide screens have told us which genes are essential for parasite survival, but knowing that a protein matters is not the same as understanding what it does. One of the most direct routes into protein function is subcellular localisation: where a protein operates inside the cell limits what it can do, and knowing its location can open the door to understanding its role.
We sought to address this by generating the first comprehensive map of protein locations inside the P. falciparum schizont, which is the stage of the parasite's life cycle in which newly formed daughter cells, each capable of invading a new red blood cell, are packed inside their host cell ready for release. Our map assigns around 1,600 proteins to one of 24 distinct cellular compartments, telling us for the first time where the majority of the parasite's proteins actually are inside the cell.
How it works
The technique we used is called hyperLOPIT (hyperplexed Localisation of Organelle Proteins by Isotopic Tagging), originally developed by our co-author Kathryn Lilley's group at the Cambridge Centre for Proteomics. Rather than studying one protein at a time, hyperLOPIT asks where all detectable proteins are located simultaneously. Cells are carefully broken open to keep internal compartments largely intact, and those compartments are then separated across a density gradient. Importantly, we are not trying to purify individual organelles at specific densities. Instead, each protein's abundance pattern across all gradient fractions provides a signature. Proteins that share the same compartment will co-disperse across the gradients and will share the same signature. By measuring these patterns for thousands of proteins using mass spectrometry, and applying machine learning to compare unknown proteins against well-characterised reference proteins, we can assign large numbers of proteins to compartments in a single experiment. In effect, we generate a protein map where all proteins are grouped according to their subcellular location and associations.
The method had been applied previously in our group to related parasites such as Toxoplasma gondii, Cryptosporidium parvum, and African trypanosomes. In those experiments, parasites were isolated from their hosts before lysis. With the schizont, we worked with parasites still inside and interacting with their host red blood cell. The parasite substantially remodels its host cell constructing entirely new compartments including Maurer's clefts and an elaborate membrane at the parasite-host interface called the parasitophorous vacuole membrane. Because these compartments resolved clearly in our maps, we could capture not just parasite organelles but also novel and remodelled host-cell niches, giving us a complete picture of the infected cell as a whole.
The rhoptry problem
One of the trickier problems we encountered involved rhoptries which are specialised secretory organelles that P. falciparum uses to invade red blood cells. The parasite injects rhoptry proteins into the host cell at the moment of entry, and these proteins contribute to the remodelled subcelluar structures of the host. In our first two experiments, rhoptry proteins refused to resolve as a single compartment on our maps. The explanation for this reflects the complexity of the parasite-infected host cell. In the schizont stage, many rhoptry proteins occupy two locations at the same time. Some sit inside newly-formed rhoptries in the daughter cells, ready for the next round of invasion. Others exist in their secreted location from the previous invasion cycle. These locations are many and protein-specific, including in the red blood cell membrane and the parasitophorous vacuole. The method was, therefore, seeing both populations at once and could not reconcile them into a single cell structure.
Our solution was to generate a third dataset from purified merozoites, which are the released invasive daughter cells. This sample, therefore, lacked the secreted rhoptry proteins from the previous host infection. Combining this dataset with the schizont data resolved the problem and the rhoptry cluster appeared on the map.
However, this three-way dataset contained around 550 fewer proteins than the two-experiment schizont dataset alone. Relying on it exclusively would mean sacrificing depth across the rest of the proteome. So we took a hybrid approach: rhoptry assignments from the three-way analysis were embedded onto the broader two-experiment map. The result was a final map of 1,646 classified proteins across 24 compartments.
Validating our own findings
A crucial part of this project was our collaboration with Julian Rayner's lab at the Cambridge Institute for Medical Research, whose expertise in P. falciparum biology shaped how we performed and interpreted every experiment. To validate our computational predictions, we worked together to engineer parasites carrying epitope tags on previously uncharacterised proteins and image them directly by microscopy. All proteins we imaged localised exactly where the map predicted, which gave us confidence in the reliability of the dataset as a whole.
What the map reveals about parasite evolution
Having location data for 1,646 proteins opened the door to questions of parasite evolution that we could not so easily have asked before. We asked, not just where proteins are, but when in evolutionary history each compartment acquired its current protein content, and which compartments are under the most intense selective pressure today.
The picture that emerged was striking. Proteins that arose recently in evolutionary terms are overwhelmingly concentrated at the interface between the parasite and the host red blood cell: the erythrocyte membrane, the Maurer's clefts, the parasitophorous vacuole membrane. These same compartments show the strongest signals of ongoing positive selection. These evolutionary signatures demonstrate the fierce interplay between parasite and host that has shaped the evolution of this organism.
The internal compartments tell a different story. The apicoplast, a relic organelle of photosynthesis that was inherited from an ancient algal ancestor, houses some of the most conserved and essential proteins in the entire parasite. Nearly everything we detected in this metabolic organelle is lethal to disrupt, making it an attractive target for drug development. However, at the level of Plasmodium speciation, apicoplast proteins show unexpected signs of positive selection, possibly reflecting metabolic adjustments as Plasmodium species adapted to infecting different vertebrate hosts over millions of years.
A resource for the community
Spatial proteomics datasets can be difficult to navigate, so alongside the paper we have developed an interactive web application where researchers can search for any protein of interest and retrieve its compartment assignment and confidence score directly. The data are also integrated into PlasmoDB, the community database for malaria parasite research, where they sit alongside the wealth of other genomic and functional data available for the parasite.
A large proportion of the P. falciparum proteome remains functionally uncharacterised. Knowing where a protein operates inside the cell offers a direct routes into understanding what it does. As the field continues to search for new drug targets and to understand how the parasite evades our interventions, we hope our data provides a foundation that the community can build on.